- From: Clark C. Evans <cce@clarkevans.com>
- Date: Mon, 22 May 2000 18:46:12 -0400 (EDT)
- To: "Simon St.Laurent" <simonstl@simonstl.com>
- cc: xml-uri@w3.org
Thank you Joe! This was *very* Helpful. On Mon, 22 May 2000, Simon St.Laurent wrote: > At 04:43 PM 5/22/00 -0400, keshlam@us.ibm.com wrote: > >OPTION 3: TREAT NAMESPACE NAMES AS LITERALS. > >^: Some of the other specs (eg XPath?) which reference Namespaces would > >need to be rewritten to reflect this understanding. Others already have > >this behavior. > > On this last bit, I'm not sure that they would need to be rewritten beyond > making explicit that their behavior goes beyond what's specified in > 'Namespaces in XML'. I really think that having "namespace" related behavior littered through other specifcations (like XPath) is a recipie for confusion, possibly even inconsistency, or worse, loss of production... A. I can have two reasonable readings of the XPath spec with regard to the namespace for "att", as <pre:el xmlns:pre="x" att="val" /> In my literal reading, the NS spec's expanded name for the attribute is { "x" , "el", "val" }, as it is in the per-element partition; however, the XPath spec talks about the namespace part and the local name part; in which case, the namespace for the attribute, is ... "x" However, James Clark implemented it as.... "" B. What happens if my XML parser follows the XML spec perfectly; and now I'm re-defining XPATH specific behavior? This leaves my SAX/DOM code different than my XPath code. This is bad news. Luckly, it is possible to build the XPath verision of namespaces on top of a good NS aware parser... however, we just have to know what to discard; for instance if the "element" is filled in for an attribute, ignore the "namespace" value. How's that for confusing? IMHO, no matter what the result here; the XPath and DOM specs should *refer* to the NS spec and the operator (identity) defined therein. It should not be in the business of building a specific version of namespaces for its own use... BTW, the blatant drop of per-element-partition knowlege from the XPath perspective makes me wonder if we shouln't axe that part of the Namespace spec while were up here debating it. I personally agree with James Clark's interpertation for the namespace of "att" in the example above. If one is really interested in the namespace of the parent element... get-namespace(parent()) Best, Clark
Received on Monday, 22 May 2000 18:43:07 UTC